"Best" means "most optimized for some utility function". It always requires the utility function to be defined, and there's always an infinite array of conflicting utility functions that could be used -- that's the nature of "best".

Stallman is, quite properly, clearly addressing the relevant utility function. This may differ from your preferred utility function, but that doesn't make RMS's definition a "redefinition of best".

How is it a "redefinition" of best? What was the original definition? Unless you add some qualifier, "best" for a software project doesn't have any kind of objective definition. For a company it might be "most profitable", for a newbie it might be "easiest to use", etc. There are a ton of criteria that can go into it. RMS helpfully explains in broad terms what in his opinion would constitute the "best Emacs".

In the context of avoiding widely desired features in order to preserve the political goals of the gnu foundation, it's straightforward to understand.